Structure and Performance of GFDL’s CM4.0 Climate Model
Observation and Context
Previous generations of climate models often suffered from persistent physical biases, including distorted ocean boundary currents, unrealistic tropical rainfall bands (the “double Intertropical Convergence Zone” or double ITCZ), and misrepresentations of tropical Pacific climate variability such as the El Niño–Southern Oscillation (ENSO). Coarse ocean resolutions often failed to capture narrow coastal currents and eddy fields, while unoptimized atmospheric tuning led to large errors in top-of-atmosphere (TOA) energy fluxes. Scientists at the Geophysical Fluid Dynamics Laboratory (GFDL) developed CM4.0 as a next-generation coupled physical model for the Coupled Model Intercomparison Project Phase 6 (CMIP6).
Hypothesis
If a climate model couples a 100-km atmospheric/land model optimized for TOA radiative balance with a high-resolution (25-km) ocean model employing hybrid vertical coordinates, it will significantly reduce global energy and precipitation biases, improve ocean boundary currents, and realistically reproduce modes of climate variability such as ENSO without requiring artificial corrections.
Experiment and Methodology
Researchers constructed and tested GFDL CM4.0 across standardized CMIP6 experimental protocols:
- Model Components: Coupled the AM4.0 atmosphere (~100-km horizontal resolution, 33 vertical levels) featuring a double-plume convection scheme and light chemistry; the LM4.0.1 land model with interactive vegetation; the MOM6 ocean model (~25-km resolution, 75 vertical hybrid z-isopycnal layers) run without mesoscale eddy parameterization; and the SIS2.0 sea ice model.
- Pre-industrial Control Simulation: Integrated a 500-year piControl run initialized from modern ocean climatologies to assess stability, internal drift, and unforced multidecadal variability.
- Historical Simulations: Ran a three-member historical ensemble (1850–2014) forced with observed greenhouse gases, solar variations, and aerosol precursor emissions to validate the model against modern satellite and reanalysis observations.
Results and Data
- Radiative Fluxes and Precipitation: CM4.0 produced lower root-mean-square errors in seasonal TOA radiative fluxes than any CMIP5 coupled model. It significantly reduced the double ITCZ bias in the tropical Pacific, achieving superior precipitation skill scores.
- Ocean Circulation and Sea Ice: The 25-km ocean resolution yielded an energetic Gulf Stream separation and a coherent Deep Western Boundary Current. Climatological Arctic sea ice extent and its recent decadal decline closely matched observations (September trend of −0.66×106 km2/decade vs. −0.87×106 km2/decade observed).
- ENSO: The simulated Niño-3 sea surface temperature power spectrum achieved a realistic multi-year period (2–8 years) and multidecadal modulation matching observational records.
- Identified Weaknesses: The model exhibited a high equilibrium climate sensitivity (~5.0 K) and strong aerosol cooling. As a result, Northern Hemisphere historical warming remained suppressed until 1990 before warming too rapidly thereafter. In the Southern Ocean, the model displayed an unrealistic ~100-year cycle of deep convective superpolynyas in the Ross Sea.
Conclusion and Climate Impact
The hypothesis was supported: coupling an atmosphere tuned for TOA energy balance with an eddy-permitting hybrid-coordinate ocean delivers a high-fidelity physical climate simulation with industry-leading skill in radiation, rainfall patterns, and ENSO. Although excessive aerosol forcing caused an unrealistic 20th-century warming trajectory and Southern Ocean variability was too strong, CM4.0 represents a major technical advance for CMIP6 and serves as a foundational baseline for modern global warming projections.
Full Citation
Held, I. M., Guo, H., Adcroft, A., Dunne, J. P., Horowitz, L. W., Krasting, J., Shevliakova, E., Winton, M., Zhao, M., Bushuk, M., Wittenberg, A. T., Wyman, B., Xiang, B., Zhang, R., Anderson, W., Balaji, V., Donner, L., Dunne, K., Durachta, J., … Zadeh, N. (2019). Structure and Performance of GFDL’s CM4.0 Climate Model. Journal of Advances in Modeling Earth Systems, 11(11), 3691–3727. https://doi.org/10.1029/2019MS001829